Complexity of scaled seismotectonic models

We investigate the complexity of two types of scaled seismotectonic models mimicking subduction megathrust seismic cycles. Our research encompasses a variety of model sizes, materials, deformation rates, and frictional configurations. Using nonlinear time-series analysis tools and displacement as an input variable, we characterize the dynamics of laboratory earthquakes in different phases of the labquake cycle. The number of active degrees of freedom that we are able to retrieve is low (<5) during most of the cycle, akin to slow slip episodes observed in natural settings and friction experiments performed with quartz powder. Results seem insensitive to the along-strike frictional segmentation of the megathrust. Nonetheless, the instantaneous dimension d can reach large values (>10), revealing the complexity of the system. High values of d correlate with slip phases, while significant drops in the extremal index anticipate slip episodes. Our results suggest that prediction horizons are in the order of a fraction of slip duration similarly to prediction horizons inferred for slow slip events in nature. This research not only enhances our understanding of earthquake dynamics, but also validates scaled seismotectonic models as effective tools for studying frictional physics across diverse spatio-temporal scales.